Upload
doanmien
View
215
Download
0
Embed Size (px)
Citation preview
Search for the Higgs boson in the4 leptons channel at ATLAS.
Nicolas Morange
CEA-IRFU (Saclay)
SLAC seminar
September 24, 2012
Starting with the conclusion
You all know the conclusion...
Discovery of a new particle in the search for the SM Higgs boson
4 leptons channel one of the major players in game
Published Phys.Lett., B716 (2012) 1-29
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 2 / 44
Starting with the conclusion
You all know the conclusion...
... which has been achieved after (only) 2.5 yearsof data-taking !
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 2 / 44
Thanks to... the LHC
Ecm:
900 GeV (2009),7 TeV (2010,2011),8 TeV (2012),13–14 TeV (2015-)
Instantaneous luminosity:
∼ 1032 cm−2s−1 (2010),∼ 1033 cm−2s−1 (2011, 2012),1034 cm−2s−1 (nominal)
Spacing between proton bunches:
50 ns (2010-2012)25 ns (nominal)
2012: 15 fb−1 and ongoing
Integrated luminosities:
2010: 45 pb−1
Day in 2011
28/02 30/04 30/06 30/08 31/10
]1
To
tal In
tegra
ted L
um
inosity [fb
0
1
2
3
4
5
6
7 = 7 TeVs ATLAS Online Luminosity
LHC Delivered
ATLAS Recorded
1Total Delivered: 5.61 fb1Total Recorded: 5.25 fb
2011: 5.25 fb−1
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 3 / 44
Thanks to... the ATLAS detectorMuon Spectrometer: (|η|< 2.7)Air toroid with drift chambers,Provides μ trigger and momentum measurement,Resolution < 10% up to p∼ 1 TeV.
Inner Detector: (|η|< 2.5, B=2T)Si Pixels, SCT, TRTPrecision tracking,Vertex reconstruction,e/π separationσ/pT ∼ 3.810−4pT⊕0.015
Hadronic Calorimeter:Scint/Fe tiles in barrel (|η|< 1.7)W/Cu-LAr in endcaps (|η|< 4.9)Provides jet trigger and energy measure-ment,σ/E∼ 50%/
pE⊕3%
Hermetic coverage for MET
EM Calorimeter: (|η|< 3.2)Pb-LAr, accordion structureProvides trigger on e/γ,Identification and measurementσ/E∼ 10%/
pE⊕0.7%
Trigger System:3 levelsL1: calo and muons, 75 kHzdedicated electronicsL2: all detectors, 4 kHzfast reconstructionEF: all detectors, 300 Hzfull reconstruction
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 4 / 44
Discovering a new particle: a long journey
2.5 years of data-taking to:
Commission and understand finelythe detector
Understand and improvereconstruction performance
Measure basic SM processes("re-discover" the SM)
Improve and perform the search forthe Higgs
ATLAS simulation 14 TeV, 2008< 5σ at 125 GeV with 10 fb−1.
All of these steps absolutely crucial to obtain thecurrent results !
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 5 / 44
Going through the steps
1 Commissionning the ATLAS detector: the L1Calo trigger system at highenergies
2 The Higgs boson at the LHC
3 Measuring basic processes: b-jets associated with Z bosons
4 Improving and performing searches: the H→ 4ℓ channel
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 6 / 44
1 Commissionning the ATLAS detector: the L1Calo trigger system at highenergies
2 The Higgs boson at the LHC
3 Measuring basic processes: b-jets associated with Z bosons
4 Improving and performing searches: the H→ 4ℓ channel
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 7 / 44
The electromagnetic calorimeter
Geometry
sampling Pb/LAr calorimeter, accordionstructure
contains EM showers
Barrel |η|< 1.47, endcaps1.37< |η|< 3.2
3 layers in depth, plus a presampler
fine transverse segmentation
⇒ ∼170 000 cells
Calorimeter Readout
Drift time 400 ns
⇒ bipolar shaping, rise time 45 ns
Optimal filtering
⇒ energy, timing, signal quality
ATLAS
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 8 / 44
The L1 calorimeter trigger system
Search for deposits compatible with physicsobjects
Computation of transverse energies,
Bunch-crossing identification (BCID).
Trigger towers
Lowered granularity: mostly∆η×∆ϕ= 0.1×0.1,
Obtained by analog sums (dedicatedelectronics, Saclay involvement),
Signals digitized at 40 MHz on 10 bits, withsteps of 250 MeV,
Energy computation: Finite Input Responsefilter (FIR ∼ optimal filtering) with 5 samples,
BCID: look for maximum of FIR output.
Trigger logic
Find Regions of Interest (local maxima),
If event accepted, transmit to L2.
nanosecond steps0 50 100 150 200 250 300 350
V
0
0.2
0.4
0.6
0.8
nanosecond steps0 50 100 150 200 250 300 350
AD
C
0
50
100
150
200
250
300
350
400
Bunch crossing0 2 4 6 8 10 12 14
[G
eV
]T
E
0
20
40
60
80
100
V
ADC
ET
ns
ns
BC
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 9 / 44
The trigger system at high energies
time [ns]
0 50 100 150 200 250
AD
C
0
200
400
600
800
1000
150 GeV
time [ns]
0 50 100 150 200 250
AD
C
0
200
400
600
800
1000
300 GeV
time [ns]
0 50 100 150 200 250
AD
C
0
200
400
600
800
1000
1.1 TeV
At very high energies:
Saturation: digital range 256 GeV. Large pulses may also be distorted
Digital saturation ⇒ L1 trigger fired. But on which bunch-crossing ?
BCID for saturated pulses:Use a threshold algorithm
1st saturated
sample
Sat. algo decision
FIR decision
L1 decision
High Threshold
Low Threshold
comparison
comparison
3953339033 34302FIR weighted sums:
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 10 / 44
Commissioning the Trigger for large pulses
Problem: initial calibration of threshold algorithm valid only for very large pulses (eg500 GeV), leaving "hole" in trigger range in 250–500 GeV
Solution: changing the trigger logic
Leave FIR algorithm on beyond 250 GeV: actually works at least up to 700 GeV
Trigger BCID: the earlier of the two outcomes
⇒ Then must prove the whole range is covered with high efficiency
time [ns]
V
Difficulties with thresholds algorithmcalibration
Sensitivity to electronics noise,
Sensitivity to digitization timings,
Shape differences between calibration andphysics signals.
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 11 / 44
Results and consequences
Validation
Uses both calibration and physics data
Makes use of the linearity of the system
Must be done for each Trigger Tower and each data-taking period
Results
BCID validated up to 3.5 TeV for most trigger towers,
Computation of uncertainties per period, for barrel and endcaps,
⇒ Application: uncertainty on W′ (1 % ) and Z′ (1.8 %) searches
Changes in trigger configuration (thresholds) for 2011 data-taking:
Better behaviour at the highest energies,⇒ Uncertainties become almost negligible.
Uncertainty on trigger efficiency fora deposit of E= 3.5 TeV (%):
5 problematic towers,
some with uncertainties∼ 5%
η and ϕ structures due to detectoreffects.
ATL-DAQ-INT-2011-001 (ATLASmembers) or CERN-THESIS-2012-087 η
2 1 0 1 2
φ
3
2
1
0
1
2
3
0
10
20
30
40
50
60
70
80
90
100
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 12 / 44
1 Commissionning the ATLAS detector: the L1Calo trigger system at highenergies
2 The Higgs boson at the LHC
3 Measuring basic processes: b-jets associated with Z bosons
4 Improving and performing searches: the H→ 4ℓ channel
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 13 / 44
The Higgs boson at the LHC
Production and decay
3 main production modes,cross-sections known to NLO or NNLO,
Many decay modes, branching ratiosknown to NLO,
Importance of different channelsdepends on hypothesised mass:ZZ, WW, γγ, V+H(bb̄), ττ,. . .
H→ ZZ: different final states, includingZZ→ 4ℓ
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 14 / 44
The Higgs to 4 leptons channel
H→ ZZ→ 4ℓ
Decays Z→ ee and μμ: 3 channels, 4e,2e2μ, 4μ
⇒ low branching ratio:σ×BR∼ 1−10 fb
Excellent resolution on MH: ∼ 2 GeV
Very clean signature, low backgrounds
⇒ Important channel on 120–600 GeV
Backgrounds
Irreducible background: ZZ
Backgrounds with leptons fromheavy-flavor decays: tt̄, Zbb̄
Backgrounds with fake leptons:Z+light jets
Zbb̄ background
Uncertainties on QCD computations of itsproduction (see W+b at Tevatron):
Multi-scales process: Q2, MZ , Mb
b-quarks mass effects
Uncertainties on processes withb-quarks in the initial state
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 15 / 44
1 Commissionning the ATLAS detector: the L1Calo trigger system at highenergies
2 The Higgs boson at the LHC
3 Measuring basic processes: b-jets associated with Z bosons
4 Improving and performing searches: the H→ 4ℓ channel
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 16 / 44
Zbb̄: b-quarks in the initial state
Origin of the issue: incompatibility between quark masses and PDF
2 different schemes
Variable number of flavors (SHERPA): use a b PDFabove threshold
Good behaviour at large Q2
Difficulties around threshold
Fixed number of flavors (ALPGEN): b created bygluon-splitting only
Needs special 4 flavors PDF.Kinematics correct around threhsoldLacks logarithmic resummations at large Q2
Xp
p
Y
X(b̄)
b
Xp
p
X
Y
b̄b
Merging the schemes
At NLO, possibility to interpolate between theschemes
Good behaviour in both limitsSmooth transitionSeveral possible prescriptions: S-ACOT(MCFM)
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 17 / 44
The Z+b analysis
Preparing the H→ 4ℓ analysis: control Zbb̄
Early data (2010) : 36pb−1 ⇒ allows measurement ofZ+b−jet
Test QCD prediction for this type of events
Background to other Higgs and SUSY searchesOn longer term, measurement of b-jets energyscale, and of b PDF.
b
b
Z
g
q Z
b
bq
In a nutshell:
Inclusive measurement on b jets
Cross-section of b-jets in Z events:
σb =N(b-jets)
LAN(b-jets) measured after b-tagging and statistical fit.
Ratio σb/σZ (number of b jets per Z event) alsomeasured
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 18 / 44
Selection of events
Standard Z selection:
Single lepton triggers (e, μ)
Good quality leptons, pT > 20 GeV, |η|< 2.5
Pair of leptons, 76<mℓℓ < 106 GeV
Adding (b-)jets:
Anti-kT , parameter 0.4, pT > 25GeV, |y|< 2.1
Separated from leptons by ∆R> 0.5
b-tagging: reconstruct a secondary vertex
Cut on the significance of the distanceto primary vertexCalibrated for 50 % efficiency
Simulations
Z+jets (signal andbackgrounds)SHERPA and ALPGEN
tt̄ MCNLO
Dibosons ALPGEN
Single top MCNLO
Simulations corrected to match dataas best as possible
[GeV]e+em60 80 100 120 140 160 180 200 220 240
Events
/ 5
GeV
110
1
10
210
310
[GeV]e+em60 80 100 120 140 160 180 200 220 240
Events
/ 5
GeV
110
1
10
210
310
= 7 TeV)sData 2010 (
) + jets
e+e→Z(
tt) + jetsττ→Z(
) + jetsνe→W(
DibosonSingle top
1 L dt = 36 pb∫
ATLAS
Invariant mass Z→ ee
Njets≥0 1 2 3 4 5
Events
1
10
210
310
410
Njets≥0 1 2 3 4 5
Events
1
10
210
310
410
= 7 TeV)sData 2010 (
) + jets
e+e→Z(
tt) + jetsττ→Z(
) + jetsνe→W(
DibosonSingle top
1 L dt = 36 pb∫
ATLAS
Number of jets (Z→ ee)
(btagged jets) [GeV]T
p20 40 60 80 100 120 140 160 180 200 220 240
Jets
/ 2
0 G
eV
110
1
10
210
(btagged jets) [GeV]T
p20 40 60 80 100 120 140 160 180 200 220 240
Jets
/ 2
0 G
eV
110
1
10
210
= 7 TeV)sData 2010 (
) + b
e+e→Z(
) + c
e+e→Z(
) + light
e+e→Z(
tt) + jetsττ→Z(
) + jetsνe→W(
DibosonSingle top
1 L dt = 36 pb∫
ATLAS
b-tagged jets spectrum(Z→ ee)
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 19 / 44
Selection of events
Standard Z selection:
Single lepton triggers (e, μ)
Good quality leptons, pT > 20 GeV, |η|< 2.5
Pair of leptons, 76<mℓℓ < 106 GeV
Adding (b-)jets:
Anti-kT , parameter 0.4, pT > 25GeV, |y|< 2.1
Separated from leptons by ∆R> 0.5
b-tagging: reconstruct a secondary vertex
Cut on the significance of the distanceto primary vertexCalibrated for 50 % efficiency
Simulations
Z+jets (signal andbackgrounds)SHERPA and ALPGEN
tt̄ MCNLO
Dibosons ALPGEN
Single top MCNLO
Simulations corrected to match dataas best as possible
[GeV]µ+µm60 80 100 120 140 160 180 200 220 240
Events
/ 5
GeV
110
1
10
210
310
[GeV]µ+µm60 80 100 120 140 160 180 200 220 240
Events
/ 5
GeV
110
1
10
210
310
= 7 TeV)sData 2010 (
) + jetsµ+µ→Z(
tt) + jetsττ→Z(
) + jetsνµ→W(
DibosonSingle top
1 L dt = 36 pb∫
ATLAS
Invariant mass Z→ μμ
Njets≥0 1 2 3 4 5
Events
1
10
210
310
410
510
Njets≥0 1 2 3 4 5
Events
1
10
210
310
410
510
= 7 TeV)sData 2010 (
) + jetsµ+µ→Z(
tt) + jetsττ→Z(
) + jetsνµ→W(
DibosonSingle top
1 L dt = 36 pb∫
ATLAS
Number of jets (Z→ μμ)
(btagged jets) [GeV]T
p20 40 60 80 100 120 140 160 180 200 220 240
Jets
/ 2
0 G
eV
110
1
10
210
(btagged jets) [GeV]T
p20 40 60 80 100 120 140 160 180 200 220 240
Jets
/ 2
0 G
eV
110
1
10
210
= 7 TeV)sData 2010 (
) + bµ+µ→Z(
) + cµ+µ→Z(
) + lightµ+µ→Z(
tt) + jetsττ→Z(
) + jetsνµ→W(
DibosonSingle top
1 L dt = 36 pb∫
ATLAS
b-tagged jets spectrum(Z→ μμ)
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 19 / 44
Results of the event selection
In data:
Selection Electron Muon≥ 1 b-tag 64 67= 1 b-tag 62 63= 2 b-tag 1 4= 3 b-tag 1 0
Data-MC agreement:
Reasonable
For both generators used
In electron and muon channels
For both yields and spectra.
Backgrounds:
Selection purity: ∼ 50%
Z+c and Z+ l dominant (taggingefficiency)
Important contribution tt̄
Dibosons, single top sub-leading
Contribution of multijets (1 event):data-driven
[GeV]e+em60 80 100 120 140 160 180 200 220 240
Events
/ 5
GeV
110
1
10
210
[GeV]e+em60 80 100 120 140 160 180 200 220 240
Events
/ 5
GeV
110
1
10
210 = 7 TeV)sData 2010 (
) + jets
e+e→Z(
tt) + jetsττ→Z(
) + jetsνe→W(
DibosonSingle top
1 L dt = 36 pb∫
ATLAS
Invariant mass Z→ ee, b-tagged jet
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 20 / 44
Extraction of the number of b-jets
Issue:
Purity not sufficient to measure cross-section directly
⇒ Need of a statistical extraction
Main backgrounds made of Z+non-b-jets
⇒ Chose invariant mass of tracks associated to the secondary vertex asdiscriminant variable
Method
Muon and electron channels added,
Templates for signal and backgroundstaken from MC,
Sub-leading backgrounds (mainly realb-jets) normalized to their estimatedcontributions,
Normalizations of signal and dominantbackgrounds are the result of the fit.
Method validated with pseudo-experiments.
SV0 mass [GeV]0 1 2 3 4 5 6 7 8 9 10
Je
ts /
0.5
Ge
V
0
10
20
30
40
50 =7TeV)sdata 2010 (Z+bZ+cZ+lightbackgrounds
SV0 mass [GeV]0 1 2 3 4 5 6 7 8 9 10
Je
ts /
0.5
Ge
V
0
10
20
30
40
50
ATLAS
1 L dt = 36 pb∫
Fit results
Nb = 63.6+14.7−13.2 Nc = 59.9+13.4
−14.0Nl = 0.0+5.1
−0.0 N(other) = 14.5 (fixed)
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 21 / 44
Systematic uncertainties
Uncertainties affect result inseveral ways
Distorsion of templates
Normalization ofsub-leading backgrounds
Change in acceptance(efficiencies)
The final result takes correlationand asymmetric uncertaintiesinto account.
Main sources:
Tagging efficiency and modeling⇒ will significantly decrease.
Source Fit (%) Acceptance (%)
Electron and MuonTagging efficiency 1.7 9.1Template modeling 3.5 -Model dependence 2.7 10.0Jet energy scale 0.7 4.0σtt̄ 2.0 -MPI negl. 1.0
Electron onlyMC statistics negl. 1.3Multijets background 1.6 -Electron efficiency negl. 5.0Total Electron 5.6 15.0
Muon onlyMC statistics negl. 1.3Multijets background 0.7 -Muon efficiency negl. 2.0Total Muon 5.4 14.3
Total uncertainty +21% -16%
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 22 / 44
Results
Measurement of the fiducial cross-section σb
Theory (MCFM): partonic NLO + hadronization corrections
Theory uncertainties: PDF, renormalization scale, αs
Experiment 3.55+0.82−0.74(stat)+0.73
−0.55(syst)±0.12(lumi) pb
MCFM 3.88±0.58pb
Published in PhysicsLetters B706:295-313,2012.
Measurement of the ratio σb/σZ:
Mean number of b jets per Z event
Interest: Direct comparison with LO generators
Experiment (7.6+1.8−1.6(stat)+1.5
−1.2(syst))×10−3
MCFM (8.8±1.1)×10−3
ALPGEN (6.20±0.2(stat only))×10−3
SHERPA (9.5±0.1(stat only))×10−3
MCFM: good agreement withmeasurement
LO generators: both at 1 sigma,large Alpgen/Sherpa difference
⇒ Advocates the use of k-factorsspecific to Z+b.
⇒ Actually used for backgroundevaluation in H→ 4ℓ december2011 search
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 23 / 44
1 Commissionning the ATLAS detector: the L1Calo trigger system at highenergies
2 The Higgs boson at the LHC
3 Measuring basic processes: b-jets associated with Z bosons
4 Improving and performing searches: the H→ 4ℓ channel
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 24 / 44
Search for H→ 4ℓ with 11fb−1 of data
Data used:Good quality data recorded by ATLAS in 2011and 2012
4.9 fb−1 for 4e in 2011,
4.8 fb−1 for 4μ and 2e2μ in 2011,
5.8 fb−1 in 2012.
High pile-up environment: up to <μ>= 30.
Principle of the search:
Select events with 2 pairs ofleptons
Reconstruct a candidate byapplying cuts on invariant masses
Further reduce backgrounds withcuts on isolations and impactparameters
Look for a peak in the 4 leptonsinvariant mass distribution, over acountinuous background made ofZZ and reducible backgrounds.
Sensitivity depends on:
Signal reconstruction efficiency
⇒ Lepton reconstrucitonefficiency
Peak width
⇒ Lepton resolution
Backgrounds rejection and control
Analysis dominated by leptonperformance
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 25 / 44
Improving the electron reconstruction
Low cross-sections involved in H4l ⇒ every increase of efficiency counts.
An improved electron reconstruction (ATLAS-CONF-2012-047)
Until 2011, track reconstruction does not take Bremsstrahlung into account⇒ loss of tracking efficiency at low pT⇒ poor estimation of track parameters.
Improved reconstruction for ATLAS, effort led by H→ 4ℓ analysis:⇒ Refit of existing electron tracks (2011-)⇒ Plus account for Bremsstrahlung possibility in track finding (2012)
Improvements of track parameters in transverse plane (d0, ϕ, q/p)
Does not change the energy measurement (calorimeter only)
Consequence on the analysis
+10 % efficiency on the signal in 4e and 2μ2e after impact parameter cuts
further gains from new tracking
What about Zbb̄ and tt̄ backgrounds ?
Non-0 impact parameters ⇒ possible bias ?
Efficiency of impact parameter cuts ?
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 26 / 44
Validation for leptons from heavy-flavor decays
Validation on simulations
Study electrons from decays of b hadrons using Z+b (ALPGEN)Electrons with pT > 7 GeV, |η|< 2.5.
Unchanged d0distribution
Significantlyimproved d0resolution
1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1
Arb
itra
ry U
nit
0
2
4
6
8
10
e (GSF)→b e (Standard)→b
ATLAS Internal=7 TeVsSimulation
d0 [mm]
1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1
GS
F/S
td
0
0.5
1
1.5
2
d0 [mm]
1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1
GS
F/S
td
0
0.5
1
1.5
21 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1
Arb
itra
ry U
nit
0
10
20
30
40
50
e (GSF)→b e (Standard)→b
Standard4=4.10µ
=0.12σ
GSF4=2.10µ
=0.09σ
ATLAS Internal=7 TeVsSimulation
d0 (recotruth) [mm]
1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1
GS
F/S
td
0
0.5
1
1.5
2
d0 (recotruth) [mm]
1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1
GS
F/S
td
0
0.5
1
1.5
2
For an unchanged cut
Slightly worserejection ofelectron from bquarks,
Far betterefficiency for Zelectrons.
0 2 4 6 8 10 12 14 16 18 20
Cut effic
iency
0
0.2
0.4
0.6
0.8
1
1.2 e (GSF)→b e (Standard)→b
ATLAS Internal=7 TeVsSimulation
(d0)σ
|d0|0 2 4 6 8 10 12 14 16 18 20
GS
F/S
td
0
0.5
1
1.5
2
(d0)σ
|d0|0 2 4 6 8 10 12 14 16 18 20
GS
F/S
td
0
0.5
1
1.5
2
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 27 / 44
Validation for leptons from heavy-flavor decays
Validation with data
Select a sample enriched with electrons from HF decays
Z+b and tt̄ (dilepton) selectionsSelect electrons near (∆R< 0.5) tagged jets490 electrons selected ; purity: ∼ 60%.
Study electron properties
(d0)σ|d0|
0 2 4 6 8 10 12 14 16 18
Entr
ies / 0
.5
0
10
20
30
40
50
60
(d0)σ|d0|
0 2 4 6 8 10 12 14 16 18
Entr
ies / 0
.5
0
10
20
30
40
50
60 Data 2011
ttZbb+jets
Z+jets
ATLAS Internal1
L dt=2.0fb∫=7 TeV, s
Standard reconstruction
(d0)σ|d0|
0 2 4 6 8 10 12 14 16 18
Entr
ies / 0
.5
0
10
20
30
40
50
60
70
(d0)σ|d0|
0 2 4 6 8 10 12 14 16 18
Entr
ies / 0
.5
0
10
20
30
40
50
60
70Data 2011
ttZbb+jets
Z+jets
ATLAS Internal1
L dt=2.0fb∫=7 TeV, s
Improved reconstruction
Results
data-MC agreement as good as with the standard reconstruction.
Improvement of the observed resolution.
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 28 / 44
December 2011 results in a nutshellPhys.Lett. B710 (2012) 383-402:
H→ 4ℓ: Observed 95 % CLs exclusion curve close to the expected one: analysis wellunderstood
Exclusion of SM Higgs: 134–156 GeV, 182–233 GeV, 256–265 GeV and268–415 GeVExclusion expected: 136–157 GeV and 184–400 GeV
H→ 4ℓ: Deviations observed around 125 GeV, 244 GeV and 500 GeV (∼ 2.1σ)
ATLAS combination: non excluded 122.5 – 129 GeV and > 539 GeVATLAS combination: excess of 2.6σ at 126 GeV
⇒ If the SM Higgs exists, it is there !
[GeV]Hm110 120 130 140 150 160 170 180
SM
σ/σ
95%
CL lim
it o
n
1
10
210
sCLObserved
sCLExpected
σ 1 ±
σ 2 ±
ATLAS
4l→(*)
ZZ→H1Ldt = 4.8 fb∫
=7 TeVs
sCLObserved
sCLExpected
σ 1 ±
σ 2 ±
Exclusion limits, low mass [GeV]Hm
200 250 300 350 400 450 500 550 600
SM
σ/σ
95%
CL lim
it o
n
1
10
210
sCLObserved
sCLExpected
σ 1 ±
σ 2 ±
ATLAS
4l→(*)
ZZ→H1Ldt = 4.8 fb∫
=7 TeVs
sCLObserved
sCLExpected
σ 1 ±
σ 2 ±
Exclusion limits, high mass [GeV]Hm
200 300 400 500 600
0Local p
510
410
310
210
110
1
ObservedExpected
ATLAS
4l→(*)
ZZ→H1Ldt = 4.8 fb∫
=7 TeVs
σ2
σ3
110
p0 values
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 29 / 44
December 2011 results in a nutshellPhys.Lett. B710 (2012) 383-402:
H→ 4ℓ: Observed 95 % CLs exclusion curve close to the expected one: analysis wellunderstood
Exclusion of SM Higgs: 134–156 GeV, 182–233 GeV, 256–265 GeV and268–415 GeVExclusion expected: 136–157 GeV and 184–400 GeV
H→ 4ℓ: Deviations observed around 125 GeV, 244 GeV and 500 GeV (∼ 2.1σ)ATLAS combination: non excluded 122.5 – 129 GeV and > 539 GeVATLAS combination: excess of 2.6σ at 126 GeV
⇒ If the SM Higgs exists, it is there !
Combined exclusion limitsCombined p0 values
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 29 / 44
Improving the H→ 4ℓ analysis
December 2011 analysis:
Very good results
Still not fully optimized for very low mass searches
⇒ Optimizations performed with simulations and 2011 data in control regions
HZ
Z
ℓ
ℓ̄
ℓ̄
ℓ
MH < 2MZ
⇒ Z off-shellAt very low masses(® 130 GeV), higher
probability for both Zoff-shell
[GeV]Z1
M0 20 40 60 80 100
[G
eV
]Z
2M
0
10
20
30
40
50
60
70
(MZ1,MZ2), MH = 120 GeV
[GeV]Z1
M0 20 40 60 80 100 120 140
[G
eV
]Z
2M
0
10
20
30
40
50
60
70
80
(MZ1,MZ2), MH = 150 GeV
Adapting to the kinematics of low-mass Higgs
Some cuts need relaxing to gain acceptance
Needs good control of backgrounds
Reasonable for Zbb̄,Z+jets less well controlled, and significant in 4e and 2μ2e.
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 30 / 44
Optimizing light jets rejection
Light jets rejection
Quite low at low pT
Study one additional shower shape cutat pT < 10 GeV
Check simulations with a study of Z+eevents
Adding the cut
Optimization in η bins
Simulations (4e and 2μ2e channels):
Signal: -2 %Backgrounds: Z+jets -50 %, Zbb̄-40 %Significance: +15 %
Eratio
0.2 0 0.2 0.4 0.6 0.8 1 1.2
Events
/ 0
.014
10
210
310
Eratio
0.2 0 0.2 0.4 0.6 0.8 1 1.2
Events
/ 0
.014
10
210
310
Data 2011Zbb+jets
Z+jets
TTbar
WZZZ
ATLAS Internal1
L dt=4.8fb∫=7 TeV, s
Eratio0.2 0 0.2 0.4 0.6 0.8 1 1.2
fraction b
elo
w c
ut
210
110
1
H (120 GeV)
Z+jets
Cut value
[0.0;0.1]∈η
ATLAS Internal=7 TeVsSimulation
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 31 / 44
Optimizing analysis cuts
Re-analyzing 2011 data
Optimization efforts carried separately for channels with subleading electrons ormuons (different backgrounds)
Here channels with subleading electrons shown
Optimizations done on simulations and control regions of 2011 data
Optimizations studied
Issue: low MC statistics for Z+jets⇒ Use control region to measure its variations
Invariant mass 1st pair of leptons: MZ−15 GeV → MZ−40 GeV
min cut [GeV]12M
40 45 50 55 60 65 70 75
Diff
ere
nce (
%)
0
20
40
60
80
H 120H 125H 130ZZZ+jets
bZbtt
ATLAS Internal=7 TeVsSimulation
min cut [GeV]12M
40 45 50 55 60 65 70 75
Diff
ere
nce (
%)
0
10
20
30
sig 120
sig 125
sig 130
ATLAS Internal=7 TeVsSimulation
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 32 / 44
Optimizing analysis cuts
Re-analyzing 2011 data
Optimization efforts carried separately for channels with subleading electrons ormuons (different backgrounds)
Here channels with subleading electrons shown
Optimizations done on simulations and control regions of 2011 data
Optimizations studied
Issue: low MC statistics for Z+jets⇒ Use control region to measure its variations
Momentum 2 highest pT leptons: 20 GeV → 15 GeV
pt min highest leptons [GeV]
15 16 17 18 19 20
Diff
ere
nce (
%)
0
10
20
30
H 120H 125H 130ZZZ+jets
bZbtt
ATLAS Internal=7 TeVsSimulation
pt min highest leptons [GeV]
15 16 17 18 19 20
Diff
ere
nce (
%)
1
0.5
0
0.5
1
sig 120
sig 125
sig 130
ATLAS Internal=7 TeVsSimulation
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 32 / 44
Optimizing analysis cuts
Re-analyzing 2011 data
Optimization efforts carried separately for channels with subleading electrons ormuons (different backgrounds)
Here channels with subleading electrons shown
Optimizations done on simulations and control regions of 2011 data
Optimizations studied
Issue: low MC statistics for Z+jets⇒ Use control region to measure its variations
Momentum 3rd lepton: 7 GeV → 10 GeV
pt min 3rd lepton [GeV]
7 8 9 10 11 12
Diff
ere
nce (
%)
100
50
0
H 120H 125H 130ZZZ+jets
bZbtt
ATLAS Internal=7 TeVsSimulation
pt min 3rd lepton [GeV]
7 8 9 10 11 12
Diff
ere
nce (
%)
0
5
10
15
20
sig 120
sig 125
sig 130
ATLAS Internal=7 TeVsSimulation
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 32 / 44
July 2012 analysis: Event selectionNumerous improvements wrt December 2011 analysis
Trigger:
Single or di-lepton triggers, thresholds evolving with luminosity
Lepton selection:
Electrons: pT > 7 GeV, |η|< 2.5, quality optimized for high efficiency
Muons: pT > 6 GeV, |η|< 2.7, maximizing the acceptance:
include spectrometer-only muons at |η|> 2.5include calorimeter-tagged muons at |η|< 0.1
Reconstruction of a candidate:
2 pairs of leptons, same flavor, opposite signs,
pT cuts at 20, 15 and 10 GeV for the 3 highest-pT leptons,
Angular separation ∆R> 0.1 for e-e and μ-μ, 0.2 for e-μ
Pair closer to the Z pole: 50<m12 < 106 GeV,
Second pair: mmin <m34 < 115 GeV.
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 33 / 44
Additional cutsGoal: suppress reducible backgrounds.
Lepton isolation Target: Z+jets, multijets
Relative isolations of tracks and clusters in cones of size ∆R< 0.2Take pileup into accountTake other leptons of the candidate into accountTrack isolation: (
∑
pT)/pℓT< 0.15
Calo isolation: (∑
ET)/EℓT< 0.30 for muons, 0.15 for standalone muons, 0.2 (2012)
or 0.3 (2011) for electrons
Transverse impact parameters Target: Zbb̄, tt̄
|d0/σ(d0)|< 3.5 (muons), 6.5 (electrons)
Efficiencies
Checked with Z→ ℓℓ decays (isolated leptons), Z+ ℓ and heavy flavor dijet events(non isolated leptons)Good data/MC agreement : ratios compatibles with 1Good simulation of pileup dependenceSmall systematic uncertainties
Final discriminant
Z mass constraint applied to leading lepton pair (m4ℓ < 190 GeV) or both(m4ℓ > 190 GeV): improvement of ∼10 % on resolution.
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 34 / 44
Signal: SimulationsGenerators:
Powheg+Pythia for gg and VBF.
Pythia for WH and ZH
Normalized to best NNLO/NLOcomputations available
Uncertainties: PDF and αs envelope(8 %), QCD scales (8 %)
Reconstruction and selectionefficiencies at 130GeV:
4μ 2e2μ 4e8 TeV data 41 % 27 % 23 %7 TeV data 43 % 23 % 17 %
december 2011 27 % 18 % 14 %
Resolutions at 130GeV:
1.8 GeV (4μ), 2.0 GeV (2e2μ), 2.4 GeV(4e)
4e: tails and lower mean value(Bremsstrahlung),
Higgs natural width dominates from350 GeV : width∼ 29 GeV àmH = 400 GeV.
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 35 / 44
Controlling backgrounds: ZZ∗
Production of SM ZZ∗:Irreducible background
Generated with Powheg (qq̄diagrams) and gg2ZZ (ggdiagrams)
Cross-section normalized to MCFMprediction
Uncertainties from QCD scales(5 %) and PDF+αs (4 % for qq̄, 8 %for gg)
Shape uncertainty from varying thegg contribution
q
q̄
Z1
Z2
ℓ̄1
ℓ1ℓ̄2
ℓ2
q
q̄
ℓ̄1
ℓ1
ℓ̄2
ℓ2q
g
gℓ̄1
ℓ1ℓ̄2
ℓ2
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 36 / 44
Controlling backgrounds: ℓℓ+μμ channels
Reducible bkg dominated by tt̄ and Zbb̄
Control region pure in thosebackgrounds: no isolation cuts,revert impact parameter cuts
Fit m12 distribution using analyticfunctions (shapes taken from MC,normalizations free)
Extrapolate to signal region usingMC ; cross-check transfer factors indata
Cross-check for tt̄: eμ+μμ controlregion
Data 4μ 2e2μZ+jets 8 TeV 0.51±0.13±0.16 0.41±0.10±0.13Z+jets 7 TeV 0.25±0.10±0.08 0.20±0.08±0.06tt̄ 8 TeV 0.044±0.015±0.015 0.040±0.013±0.013tt̄ 7 TeV 0.022±0.010±0.011 0.020±0.009±0.011
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 37 / 44
Controlling backgrounds: ℓℓ+ee channels
Reducible bkg dominated by Z+jets
Control region: relax identificationof subleading electrons. ∼100events per year and per channel
Categorize electrons inelectron-like, conversion-like andfake-like using info from trackerand calorimeter (9 categories)
Extrapolate contamination in signalregion from these categories usingMC
Two other methods used forcross-checks
Data 4e 2μ2eBackground 8 TeV 3.9±0.7±0.8 4.9±0.8±0.7Background 7 TeV 3.1±0.6±0.5 2.6±0.4±0.4
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 38 / 44
Systematic uncertainties
Normalizations:
Effects of PDF, αs and QCD scales,
Additional uncertainty for high mass searches: 150%×m3H[TeV] for mH > 300 GeV,
Luminosity: 3.6 % (2012), 1.8 % (2011).
Reducible backgrounds:
ℓℓ+μμ channels: 50 %,
ℓℓ+ee channels: 25 %,Origin: Statistical uncertainties in control regions, uncertainties on the methodsthemselves and on the extrapolations
Reconstruction and selection efficiencies:
Muon efficiency: 0.16 % (4μ), 0.12 % (2e2μ),
Electron efficiency: 3.0 % (4e), 1.7 % (2e2μ) at 600 GeV. 8.0 % and 4.6 % at 110 GeV,
Lepton resolution: negligible,
Electron energy scale: uncertainty 0.7 % (4e) and 0.4 % (2e2μ) on mass scale.
Isolations and impact parameters: negligible
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 39 / 44
Systematic uncertainties
Normalizations:
Effects of PDF, αs and QCD scales,
Additional uncertainty for high mass searches: 150%×m3H[TeV] for mH > 300 GeV,
Luminosity: 3.6 % (2012), 1.8 % (2011).
Reducible backgrounds:
ℓℓ+μμ channels: 50 %,
ℓℓ+ee channels: 25 %,Origin: Statistical uncertainties in control regions, uncertainties on the methodsthemselves and on the extrapolations
Reconstruction and selection efficiencies:
Muon efficiency: 0.16 % (4μ), 0.12 % (2e2μ),
Electron efficiency: 3.0 % (4e), 1.7 % (2e2μ) at 600 GeV. 8.0 % and 4.6 % at 110 GeV,
Lepton resolution: negligible,
Electron energy scale: uncertainty 0.7 % (4e) and 0.4 % (2e2μ) on mass scale.
Isolations and impact parameters: negligible
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 39 / 44
Systematic uncertainties
Normalizations:
Effects of PDF, αs and QCD scales,
Additional uncertainty for high mass searches: 150%×m3H[TeV] for mH > 300 GeV,
Luminosity: 3.6 % (2012), 1.8 % (2011).
Reducible backgrounds:
ℓℓ+μμ channels: 50 %,
ℓℓ+ee channels: 25 %,Origin: Statistical uncertainties in control regions, uncertainties on the methodsthemselves and on the extrapolations
Reconstruction and selection efficiencies:
Muon efficiency: 0.16 % (4μ), 0.12 % (2e2μ),
Electron efficiency: 3.0 % (4e), 1.7 % (2e2μ) at 600 GeV. 8.0 % and 4.6 % at 110 GeV,
Lepton resolution: negligible,
Electron energy scale: uncertainty 0.7 % (4e) and 0.4 % (2e2μ) on mass scale.
Isolations and impact parameters: negligible
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 39 / 44
Results: Phys. Lett. B 716 (2012) 1-29
Events in window 120–130GeV
Signal ZZ(∗) Z+ jets, tt̄ Observed4μ 2.09±0.30 1.12±0.05 0.13±0.04 6
2e2μ/2μ2e 2.29± 0.33 0.80±0.05 1.27±0.19 54e 0.90±0.14 0.44±0.04 1.09±0.20 2
Reducible backgrounds still large in 4e/2μ2e channels
Data exceed background expectations around 125 GeV
Invariant mass distribution Masses of the reconstructed Z pairs
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 40 / 44
Event display
4μ candidate. m4ℓ = 123.5 GeV.Masses of the lepton pairs: 84 GeV and 45.7 GeV.
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 41 / 44
Investigation of the excess
Test statistic: maximum likelihood fit to data using signal and background histograms.
H→ 4ℓ Exclusion limits set with CLs formalism:
Expected 124–164 GeV and 176–500 GeVObserved 131–162 GeV and 170–460 GeV
One significant excess in mass range: p0 of 3.6σ at 125 GeV (2.7 expected)
⇒ Evidence for new particleWith look-elsewhere effect in 110–150: global p0 of 2.5σ
ATLAS combination: observation at 6.0σ of a new particle of mass 126.0±0.6 GeV
signal strength parameters compatible with 14ℓ and γγ masses compatible
H→ 4ℓ low-mass exclusion limits H→ 4ℓ high-mass exclusion limits
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 42 / 44
Investigation of the excess
Test statistic: maximum likelihood fit to data using signal and background histograms.
H→ 4ℓ Exclusion limits set with CLs formalism:
Expected 124–164 GeV and 176–500 GeVObserved 131–162 GeV and 170–460 GeV
One significant excess in mass range: p0 of 3.6σ at 125 GeV (2.7 expected)
⇒ Evidence for new particleWith look-elsewhere effect in 110–150: global p0 of 2.5σ
ATLAS combination: observation at 6.0σ of a new particle of mass 126.0±0.6 GeV
signal strength parameters compatible with 14ℓ and γγ masses compatible
H→ 4ℓ p0 values H→ 4ℓ (μ,mH) contour
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 42 / 44
Investigation of the excess
Test statistic: maximum likelihood fit to data using signal and background histograms.
H→ 4ℓ Exclusion limits set with CLs formalism:
Expected 124–164 GeV and 176–500 GeVObserved 131–162 GeV and 170–460 GeV
One significant excess in mass range: p0 of 3.6σ at 125 GeV (2.7 expected)
⇒ Evidence for new particleWith look-elsewhere effect in 110–150: global p0 of 2.5σ
ATLAS combination: observation at 6.0σ of a new particle of mass 126.0±0.6 GeV
signal strength parameters compatible with 14ℓ and γγ masses compatible
Combined p0 values Superposition of (μ,mH) contours
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 42 / 44
What next ?
Have we discovered the SM Higgs ?
Goal: Test compatibility of SM oralternative theories to themeasurements
What can we achieve ?
Measurement of mass
Determination of spin
Determination of parity
Measurement of the couplings
How ?
Go for precision measurements inthe "discovery" channels
Look for a signal in other modes:H→ bb̄, H→ ττ, tt̄H of primeimportance
Look for exclusive signatures: VBF,associated production
Lot of work ahead !
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 43 / 44
Conclusions
Discovery of a new Higgs-like particle : a major result
It took 20 years of preparation, and 2.5 years of data-taking to achieve it
Commissioning ATLAS: high energy pulses at L1
An important part of the detector to get ready at the start of data-takingDifficult validation, which led to changes in trigger logicDirect consequences for some exotic searches
Prepare for the H→ 4ℓ search: Z+b measurement with 2010 data
Measurement with 30 % uncertainty, in agreement with NLOGives confidence in evaluation of Zbb̄ background in the H→ 4ℓ channel:computation of k-factor applied on ALPGEN samples
Searching for the Higgs: the 4 leptons channel
Lepton performance of prime importanceDevelopement of a new electron reconstruction
After december 2011 results, series of optimizations for low-mass searchesImprovements on background rejection and optimized kinematical cuts
Search with 2011 and 2012 data: 3.6σ excess at 125.0 GeVCompatible with SM Higgs and the excesses in other channelsA new field now opens: precision tests of SM and alternative theories
All decay channels should be studied to obtain the best possible precisionSoon 20fb−1 available and improved analyses in the high resolution channels
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 44 / 44
Basics of validation of the settings
Goal
Determine energy ranges where BCID algorithms are valid,
Compute lower bounds on trigger efficiency at high energy.
Constraints
Should be done for each trigger tower,
Should be done for each data-taking period,
Cannot rely on calibration data only,
Very small statistics of high energy deposits inphysics data.
Solution
Use calibration data to detect hardware issues,
Use calibration data to understand behaviour ofsignals at high energies,
Make use of signal linearities to extrapolate lowenergy physics data up to high energies.
TE
0 500 1000 1500 2000 2500 3000
AD
C
0
200
400
600
800
1000
Calibration
ADC(n)50 100 150 200 250
AD
C(n
1)
20
40
60
80
100
120
Physics
ET
ADC(n-1)
ADC(n-1)
ADC(n)
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 2 / 4
Estimation of multijets background
Data-driven estimateAssume an exponential invariant mass spectrum for this background.
Relax electron identification: 1 medium 1 candidate, or 2 loose no medium(cross-check)
Fit m(Z): function for multijets, MC distributions for signal and other backgrounds
Fit standard selection, with expo. slope from previous fit.
Extract number of events under Z peak: 1.0±2.2
ee mass [GeV]60 70 80 90 100 110 120
Eve
nts
/ (
3 )
0
10
20
30
40
50
60
39±nmc = 133
43±nqcd = 463
0.0033±slope = 0.02283
ee mass [GeV]60 70 80 90 100 110 120
Eve
nts
/ (
3 )
0
10
20
30
40
50
60
1L=36.2pb∫
Extraction of slope, cand+medium
ee mass [GeV]60 70 80 90 100 110 120
Eve
nts
/ (
3 )
0
5
10
15
20
25
30
9.8±nmc = 71.8
5.1±nqcd = 2.2
ee mass [GeV]60 70 80 90 100 110 120
Eve
nts
/ (
3 )
0
5
10
15
20
25
30
1L=36.2pb∫
Fit medium+medium
Muons channel: Use non isolated muon pairs. Result: N = 0.0±0.9.
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 3 / 4
Cross-section computation
Measurement: mean cross-section perlepton flavor
σb =Nb
AeLe +AμLμ
Nb: fit result
A: from simulation. Take into account:
Tagging efficiency,Jets reconstruction,Lepton reconstructionSmall extrapolation of leptonsacceptance.
ALPGEN and SHERPA numberscompatible.Aμ = 0.286, Ae = 0.214
L: luminosity: 36 pb−1.
Fiducial volume
Truth-level selections
Z boson:
Two leptons pT > 20 GeV, |η|< 2.5
Invariant mass in MZ±15 GeV
Jets:
Reconstrcuted from all final stateparticles, Z leptons excepted.
pT > 25 GeV, |y|< 2.1
Separated from Z leptons
Presence of a B hadron withpT > 5 GeV within ∆R< 0.3
Nicolas Morange (IRFU) SLAC seminar 24/09/2012 4 / 4